Title: Rescaling interval regression with spike and truncation point
Dear all,
I am trying to estimate an interval regression model, where the dependent variable is weekly number of hours (H). With four intervals: 0 to 10, 11 to 20, 21 to 40, 41 or more.
In reality, though, the range of the H goes from zero to 168 (total hours of a week).
For this reason I would like to truncate H, with a spike at zero and at a maximum point of 168.
In order to get correct estimates I have to rescale the probabilities.
However, I don't know how to do this in practical terms.
Does anyone have any suggestion please?
Also, is there any difference in the rescaling when I estimate the same model using ordered probit model?
Finally, if I need to use the log of H could I log the interval points and for the lower bound replace the missing value (missing because the log of zero is returned as missing for the lowest bound) with zero and then go on with rescaling the probabilities?
Thank you very much
Manos
[email protected]
